X-Git-Url: https://git.carlh.net/gitweb/?a=blobdiff_plain;f=src%2Flib%2Fdecoder.cc;h=e154aac5d10b3f3fd2d3b878798eb776a7337afc;hb=09a95eca971ddd2ea75cd615e7ef2b997d52b690;hp=a90c14b2bdc995af6664e51965a68bd7995978f8;hpb=c0e04acd1e9875fa67800a7861bd8a370157b49f;p=dcpomatic.git diff --git a/src/lib/decoder.cc b/src/lib/decoder.cc index a90c14b2b..e154aac5d 100644 --- a/src/lib/decoder.cc +++ b/src/lib/decoder.cc @@ -24,60 +24,45 @@ #include #include #include -extern "C" { -#include -#include -#if (LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 53 && LIBAVFILTER_VERSION_MINOR <= 77) || LIBAVFILTER_VERSION_MAJOR == 3 -#include -#include -#elif LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15 -#include -#endif -#include -} #include "film.h" #include "format.h" #include "job.h" -#include "film_state.h" #include "options.h" #include "exceptions.h" #include "image.h" #include "util.h" #include "log.h" #include "decoder.h" -#include "filter.h" #include "delay_line.h" -#include "ffmpeg_compatibility.h" #include "subtitle.h" +#include "filter_graph.h" -using namespace std; -using namespace boost; +using std::string; +using std::stringstream; +using std::min; +using std::list; +using boost::shared_ptr; -/** @param s FilmState of the Film. +/** @param f Film. * @param o Options. * @param j Job that we are running within, or 0 - * @param l Log to use. * @param minimal true to do the bare minimum of work; just run through the content. Useful for acquiring * accurate frame counts as quickly as possible. This generates no video or audio output. * @param ignore_length Ignore the content's claimed length when computing progress. */ -Decoder::Decoder (boost::shared_ptr s, boost::shared_ptr o, Job* j, Log* l, bool minimal, bool ignore_length) - : _fs (s) +Decoder::Decoder (boost::shared_ptr f, boost::shared_ptr o, Job* j, bool minimal, bool ignore_length) + : _film (f) , _opt (o) , _job (j) - , _log (l) , _minimal (minimal) , _ignore_length (ignore_length) - , _video_frame (0) - , _buffer_src_context (0) - , _buffer_sink_context (0) - , _have_setup_video_filters (false) + , _video_frame_index (0) , _delay_line (0) , _delay_in_bytes (0) , _audio_frames_processed (0) { - if (_opt->decode_video_frequency != 0 && _fs->length() == 0) { - throw DecodeError ("cannot do a partial decode if length == 0"); + if (_opt->decode_video_frequency != 0 && !_film->length()) { + throw DecodeError ("cannot do a partial decode if length is unknown"); } } @@ -90,7 +75,7 @@ Decoder::~Decoder () void Decoder::process_begin () { - _delay_in_bytes = _fs->audio_delay() * _fs->audio_sample_rate() * _fs->audio_channels() * bytes_per_audio_sample() / 1000; + _delay_in_bytes = _film->audio_delay() * audio_sample_rate() * audio_channels() * bytes_per_audio_sample() / 1000; delete _delay_line; _delay_line = new DelayLine (_delay_in_bytes); @@ -104,7 +89,7 @@ Decoder::process_end () if (_delay_in_bytes < 0) { uint8_t remainder[-_delay_in_bytes]; _delay_line->get_remaining (remainder); - _audio_frames_processed += _delay_in_bytes / (_fs->audio_channels() * bytes_per_audio_sample()); + _audio_frames_processed += _delay_in_bytes / (audio_channels() * bytes_per_audio_sample()); emit_audio (remainder, -_delay_in_bytes); } @@ -112,24 +97,28 @@ Decoder::process_end () in to get it to the right length. */ - int64_t const video_length_in_audio_frames = ((int64_t) _fs->dcp_length() * _fs->target_sample_rate() / _fs->frames_per_second()); + int64_t const video_length_in_audio_frames = ((int64_t) video_frame_index() * audio_sample_rate() / frames_per_second()); int64_t const audio_short_by_frames = video_length_in_audio_frames - _audio_frames_processed; - _log->log ( + _film->log()->log ( String::compose ("DCP length is %1 (%2 audio frames); %3 frames of audio processed.", - _fs->dcp_length(), + video_frame_index(), video_length_in_audio_frames, _audio_frames_processed) ); if (audio_short_by_frames >= 0 && _opt->decode_audio) { - _log->log (String::compose ("DCP length is %1; %2 frames of audio processed.", _fs->dcp_length(), _audio_frames_processed)); - _log->log (String::compose ("Adding %1 frames of silence to the end.", audio_short_by_frames)); + _film->log()->log (String::compose ("DCP length is %1; %2 frames of audio processed.", video_frame_index(), _audio_frames_processed)); + _film->log()->log (String::compose ("Adding %1 frames of silence to the end.", audio_short_by_frames)); + + /* XXX: this is slightly questionable; does memset () give silence with all + sample formats? + */ - int64_t bytes = audio_short_by_frames * _fs->audio_channels() * bytes_per_audio_sample(); + int64_t bytes = audio_short_by_frames * _film->audio_channels() * bytes_per_audio_sample(); - int64_t const silence_size = 16 * 1024 * _fs->audio_channels() * bytes_per_audio_sample(); + int64_t const silence_size = 16 * 1024 * _film->audio_channels() * bytes_per_audio_sample(); uint8_t silence[silence_size]; memset (silence, 0, silence_size); @@ -147,13 +136,13 @@ Decoder::go () { process_begin (); - if (_job && _ignore_length) { + if (_job && !_film->dcp_length()) { _job->set_progress_unknown (); } while (pass () == false) { - if (_job && !_ignore_length) { - _job->set_progress (float (_video_frame) / _fs->dcp_length()); + if (_job && _film->dcp_length()) { + _job->set_progress (float (_video_frame_index) / _film->dcp_length().get()); } } @@ -167,12 +156,7 @@ Decoder::go () bool Decoder::pass () { - if (!_have_setup_video_filters) { - setup_video_filters (); - _have_setup_video_filters = true; - } - - if (!_ignore_length && _video_frame >= _fs->dcp_length()) { + if (!_ignore_length && _video_frame_index >= _film->dcp_length()) { return true; } @@ -180,7 +164,7 @@ Decoder::pass () } /** Called by subclasses to tell the world that some audio data is ready - * @param data Audio data, in FilmState::audio_sample_format. + * @param data Audio data, in Film::audio_sample_format. * @param size Number of bytes of data. */ void @@ -195,13 +179,20 @@ Decoder::process_audio (uint8_t* data, int size) void Decoder::emit_audio (uint8_t* data, int size) { + if (size == 0) { + return; + } + + assert (_film->audio_channels()); + assert (bytes_per_audio_sample()); + /* Deinterleave and convert to float */ - assert ((size % (bytes_per_audio_sample() * _fs->audio_channels())) == 0); + assert ((size % (bytes_per_audio_sample() * audio_channels())) == 0); int const total_samples = size / bytes_per_audio_sample(); - int const frames = total_samples / _fs->audio_channels(); - shared_ptr audio (new AudioBuffers (_fs->audio_channels(), frames)); + int const frames = total_samples / _film->audio_channels(); + shared_ptr audio (new AudioBuffers (audio_channels(), frames)); switch (audio_sample_format()) { case AV_SAMPLE_FMT_S16: @@ -213,7 +204,7 @@ Decoder::emit_audio (uint8_t* data, int size) audio->data(channel)[sample] = float(*p++) / (1 << 15); ++channel; - if (channel == _fs->audio_channels()) { + if (channel == _film->audio_channels()) { channel = 0; ++sample; } @@ -230,7 +221,7 @@ Decoder::emit_audio (uint8_t* data, int size) audio->data(channel)[sample] = float(*p++) / (1 << 31); ++channel; - if (channel == _fs->audio_channels()) { + if (channel == _film->audio_channels()) { channel = 0; ++sample; } @@ -240,7 +231,7 @@ Decoder::emit_audio (uint8_t* data, int size) case AV_SAMPLE_FMT_FLTP: { float* p = reinterpret_cast (data); - for (int i = 0; i < _fs->audio_channels(); ++i) { + for (int i = 0; i < _film->audio_channels(); ++i) { memcpy (audio->data(i), p, frames * sizeof(float)); p += frames; } @@ -252,9 +243,9 @@ Decoder::emit_audio (uint8_t* data, int size) } /* Maybe apply gain */ - if (_fs->audio_gain() != 0) { - float const linear_gain = pow (10, _fs->audio_gain() / 20); - for (int i = 0; i < _fs->audio_channels(); ++i) { + if (_film->audio_gain() != 0) { + float const linear_gain = pow (10, _film->audio_gain() / 20); + for (int i = 0; i < _film->audio_channels(); ++i) { for (int j = 0; j < frames; ++j) { audio->data(i)[j] *= linear_gain; } @@ -262,7 +253,7 @@ Decoder::emit_audio (uint8_t* data, int size) } /* Update the number of audio frames we've pushed to the encoder */ - _audio_frames_processed += frames; + _audio_frames_processed += audio->frames (); Audio (audio); } @@ -275,183 +266,63 @@ void Decoder::process_video (AVFrame* frame) { if (_minimal) { - ++_video_frame; + ++_video_frame_index; return; } - /* Use FilmState::length here as our one may be wrong */ + /* Use Film::length here as our one may be wrong */ int gap = 0; if (_opt->decode_video_frequency != 0) { - gap = _fs->length() / _opt->decode_video_frequency; + gap = _film->length().get() / _opt->decode_video_frequency; } - if (_opt->decode_video_frequency != 0 && gap != 0 && (_video_frame % gap) != 0) { - ++_video_frame; + if (_opt->decode_video_frequency != 0 && gap != 0 && (_video_frame_index % gap) != 0) { + ++_video_frame_index; return; } -#if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 53 && LIBAVFILTER_VERSION_MINOR <= 61 + shared_ptr graph; - if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0) < 0) { - throw DecodeError ("could not push buffer into filter chain."); + list >::iterator i = _filter_graphs.begin(); + while (i != _filter_graphs.end() && !(*i)->can_process (Size (frame->width, frame->height), (AVPixelFormat) frame->format)) { + ++i; } -#elif LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15 - - AVRational par; - par.num = sample_aspect_ratio_numerator (); - par.den = sample_aspect_ratio_denominator (); - - if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0, par) < 0) { - throw DecodeError ("could not push buffer into filter chain."); - } - -#else - - if (av_buffersrc_write_frame (_buffer_src_context, frame) < 0) { - throw DecodeError ("could not push buffer into filter chain."); + if (i == _filter_graphs.end ()) { + graph.reset (new FilterGraph (_film, this, _opt->apply_crop, Size (frame->width, frame->height), (AVPixelFormat) frame->format)); + _filter_graphs.push_back (graph); + _film->log()->log (String::compose ("New graph for %1x%2, pixel format %3", frame->width, frame->height, frame->format)); + } else { + graph = *i; } -#endif - -#if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 15 && LIBAVFILTER_VERSION_MINOR <= 61 - while (avfilter_poll_frame (_buffer_sink_context->inputs[0])) { -#else - while (av_buffersink_read (_buffer_sink_context, 0)) { -#endif + list > images = graph->process (frame); -#if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 15 - - int r = avfilter_request_frame (_buffer_sink_context->inputs[0]); - if (r < 0) { - throw DecodeError ("could not request filtered frame"); + for (list >::iterator i = images.begin(); i != images.end(); ++i) { + if (_opt->black_after > 0 && _video_frame_index > _opt->black_after) { + (*i)->make_black (); } - AVFilterBufferRef* filter_buffer = _buffer_sink_context->inputs[0]->cur_buf; - -#else - - AVFilterBufferRef* filter_buffer; - if (av_buffersink_get_buffer_ref (_buffer_sink_context, &filter_buffer, 0) < 0) { - filter_buffer = 0; + shared_ptr sub; + if (_timed_subtitle && _timed_subtitle->displayed_at (double (video_frame_index()) / _film->frames_per_second())) { + sub = _timed_subtitle->subtitle (); } - -#endif - if (filter_buffer) { - /* This takes ownership of filter_buffer */ - shared_ptr image (new FilterBufferImage ((PixelFormat) frame->format, filter_buffer)); - - if (_opt->black_after > 0 && _video_frame > _opt->black_after) { - image->make_black (); - } - - shared_ptr sub; - if (_timed_subtitle && _timed_subtitle->displayed_at (double (last_video_frame()) / rint (_fs->frames_per_second()))) { - sub = _timed_subtitle->subtitle (); - } - - TIMING ("Decoder emits %1", _video_frame); - Video (image, _video_frame, sub); - ++_video_frame; - } + TIMING ("Decoder emits %1", _video_frame_index); + Video (*i, _video_frame_index, sub); + ++_video_frame_index; + _last_image = *i; + _last_subtitle = sub; } } - -/** Set up a video filtering chain to include cropping and any filters that are specified - * by the Film. - */ void -Decoder::setup_video_filters () +Decoder::repeat_last_video () { - stringstream fs; - Size size_after_crop; - - if (_opt->apply_crop) { - size_after_crop = _fs->cropped_size (native_size ()); - fs << crop_string (Position (_fs->crop().left, _fs->crop().top), size_after_crop); - } else { - size_after_crop = native_size (); - fs << crop_string (Position (0, 0), size_after_crop); - } - - string filters = Filter::ffmpeg_strings (_fs->filters()).first; - if (!filters.empty ()) { - filters += ","; - } - - filters += fs.str (); - - avfilter_register_all (); - - AVFilterGraph* graph = avfilter_graph_alloc(); - if (graph == 0) { - throw DecodeError ("Could not create filter graph."); - } - - AVFilter* buffer_src = avfilter_get_by_name("buffer"); - if (buffer_src == 0) { - throw DecodeError ("Could not find buffer src filter"); - } - - AVFilter* buffer_sink = get_sink (); - - stringstream a; - a << native_size().width << ":" - << native_size().height << ":" - << pixel_format() << ":" - << time_base_numerator() << ":" - << time_base_denominator() << ":" - << sample_aspect_ratio_numerator() << ":" - << sample_aspect_ratio_denominator(); - - int r; - - if ((r = avfilter_graph_create_filter (&_buffer_src_context, buffer_src, "in", a.str().c_str(), 0, graph)) < 0) { - throw DecodeError ("could not create buffer source"); - } - - AVBufferSinkParams* sink_params = av_buffersink_params_alloc (); - PixelFormat* pixel_fmts = new PixelFormat[2]; - pixel_fmts[0] = pixel_format (); - pixel_fmts[1] = PIX_FMT_NONE; - sink_params->pixel_fmts = pixel_fmts; - - if (avfilter_graph_create_filter (&_buffer_sink_context, buffer_sink, "out", 0, sink_params, graph) < 0) { - throw DecodeError ("could not create buffer sink."); - } - - AVFilterInOut* outputs = avfilter_inout_alloc (); - outputs->name = av_strdup("in"); - outputs->filter_ctx = _buffer_src_context; - outputs->pad_idx = 0; - outputs->next = 0; - - AVFilterInOut* inputs = avfilter_inout_alloc (); - inputs->name = av_strdup("out"); - inputs->filter_ctx = _buffer_sink_context; - inputs->pad_idx = 0; - inputs->next = 0; - - _log->log ("Using filter chain `" + filters + "'"); - -#if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15 - if (avfilter_graph_parse (graph, filters.c_str(), inputs, outputs, 0) < 0) { - throw DecodeError ("could not set up filter graph."); - } -#else - if (avfilter_graph_parse (graph, filters.c_str(), &inputs, &outputs, 0) < 0) { - throw DecodeError ("could not set up filter graph."); - } -#endif - - if (avfilter_graph_config (graph, 0) < 0) { - throw DecodeError ("could not configure filter graph."); - } - - /* XXX: leaking `inputs' / `outputs' ? */ + assert (_last_image); + Video (_last_image, _video_frame_index, _last_subtitle); + ++_video_frame_index; } void @@ -459,9 +330,9 @@ Decoder::process_subtitle (shared_ptr s) { _timed_subtitle = s; - if (_opt->apply_crop) { + if (_timed_subtitle && _opt->apply_crop) { Position const p = _timed_subtitle->subtitle()->position (); - _timed_subtitle->subtitle()->set_position (Position (p.x - _fs->crop().left, p.y - _fs->crop().top)); + _timed_subtitle->subtitle()->set_position (Position (p.x - _film->crop().left, p.y - _film->crop().top)); } }